How does this improve AI code quality?

It stops the AI from assuming code works based on confidence and forces it to prove success through actual test outputs, linter results, and build exit codes.

Can this be used for regression testing?

Absolutely. It includes a specific pattern for TDD Red-Green cycles, ensuring a bug is actually reproducible before it is claimed as fixed.

What is the 'Iron Law' in this skill?

The Iron Law dictates that no completion claims can be made without fresh verification evidence. If a verification command hasn't been run in the current context, the task cannot be called 'done'.

Does it prevent common AI mistakes like 'hallucinated' success?

Yes. It specifically targets 'red flag' language like 'should work' or 'seems correct' and replaces it with a requirement for raw command evidence.

Is this skill compatible with existing CI/CD workflows?

Yes, it acts as a local pre-flight check, ensuring that Claude only pushes code that is already verified to pass the same checks your CI/CD pipeline would run.

Verified Completion Guard

Name: Verified Completion Guard
Author: as4584

byas4584

0•

セキュリティとテスト

Enforces rigorous verification and evidence-based reporting before claiming task completion or submitting code changes.

The Verified Completion Guard implements a strict 'evidence-first' protocol for Claude Code, preventing premature or dishonest completion claims. It mandates the execution of fresh verification commands—such as tests, builds, and linters—and requires the agent to confirm success through actual command output rather than assumptions. By establishing a non-negotiable gate function before any PR, commit, or status update, it ensures high reliability, maintains developer trust, and prevents the shipment of broken or incomplete features.

主な機能

01Real-time exit code and failure count analysis

02Comprehensive 'Red-Green' verification for regression testing

03Rationalization prevention against 'should' or 'probably' language

04Mandatory command execution before any success claims

050 GitHub stars

06Independent verification of delegated agent reports and VCS diffs

ユースケース

01Verifying complex feature requirements against a line-by-line checklist

02Ensuring clean builds and linter passes before creating pull requests

03Validating bug fixes through reproduction and resolution testing cycles

What are Skills?·How to Install

Install with 🐟 Skill.Fish

npx skillfish add as4584/antigravity-skills verification-before-completion

For use in Claude.ai and ChatGPT

Download Skill